Loan Automation

Loan processing within legal practices is notoriously labor-intensive: case workers manually transcribe customer records from spreadsheets, draft individualized legal agreements, and dispatch repetitive status updates across disparate communication channels. Loan Automation was engineered as a comprehensive backend orchestration platform that eliminates manual intervention entirely.

As the full-stack automation developer, I designed the complete system lifecycle from initial spreadsheet ingestion to production deployment on a Linux VPS. The platform transforms raw customer datasets into structured JSON models, feeds those models into dynamic PDF compilation services, triggers multi-channel messaging via WhatsApp (Interakt) and email (ZeptoMail), and synchronizes workflow states through bidirectional webhooks.

The entry point of the automation workflow begins with ingesting client loan spreadsheets. In legal environments, spreadsheet files often contain subtle formatting discrepancies, missing fields, or irregular column arrangements. I implemented a robust file-intake interface allowing staff to select and upload Excel workbooks with instant client-side feedback.

Upon file selection, the backend validates file extensions, MIME types, and header structures before streaming the contents into our transformation pipeline. This pre-flight validation guarantees that corrupt files or incomplete records are identified before entering downstream legal document generation.

To bridge the gap between human-maintained spreadsheets and machine-readable services, I built an automated Excel-to-JSON transformation engine. The parser processes uploaded workbooks row-by-row, extracting key financial terms, customer contact coordinates, interest schedules, and legal identifiers.

The engine enforces strict schema normalization: date formats are converted into standardized ISO strings, monetary values are parsed into clean numeric types, and customer phone numbers are sanitized for international WhatsApp dispatching. The resulting structured JSON payload forms a dependable, clean contract consumed by the rest of the application.

Following successful JSON normalization, the extracted loan data is rendered in a clean, high-visibility operations table. This dashboard allows legal administrators to inspect ingested records, review extracted customer parameters, and audit loan variables before initiating automated processing batches.

Each record displays borrower identification, loan amounts, approval stages, and communication readiness. Administrators maintain full oversight over which records are queued for document compilation and automated dispatch, providing an essential balance between full automation and human auditability.

Legal agreements require rigorous typography, precise page geometry, and dynamic variable binding. Rather than relying on manual template filling, I developed an automated PDF generation service capable of synthesizing complete loan contracts programmatically.

Using headless document generation pipelines, the service takes normalized JSON record data and injects borrower names, loan principals, interest schedules, and repayment timelines into formal legal templates. The resulting PDF documents feature crisp vector typography, proper pagination, and embedded legal disclaimers, producing production-ready contracts in seconds.

Delivering loan agreements and updates promptly to borrowers is paramount. I engineered a multi-channel communication layer integrating Interakt for WhatsApp Business messaging and ZeptoMail for high-deliverability transactional email.

Once a borrower's PDF agreement is compiled, the backend automatically dispatches personalized WhatsApp notifications through Interakt, attaching the document link and custom greeting. Simultaneously, ZeptoMail delivers a formal email confirmation with the signed contract attached. This dual-channel approach achieves near-100% open rates and immediate borrower acknowledgment without requiring manual staff outreach.

A business-critical automation platform requires an infrastructure engineered for persistent uptime. I managed the complete deployment of the system onto a production Linux Virtual Private Server (VPS), configuring the entire server software stack from the ground up.

I configured Nginx as a high-performance reverse proxy handling SSL/TLS termination, request rate limiting, and HTTP-to-HTTPS redirects. To ensure the Node.js application runs continuously without interruption, I configured PM2 process clustering with automated restarts upon server reboots, complemented by Let's Encrypt SSL certificates securing all incoming webhook payloads and API traffic.